[Appeared in: Information Center Magazine, March 1991.]
As anyone who has spent an hour vainly looking through a manual for the answer to a question would testify, information is not knowledge. New systems and products arrive with several manuals and a troubleshooting guide - supposedly all the information you need to support your users. But you still can't answer their questions until you look up the answer, or call for help yourself. Modern knowledge distribution systems, which apply expert systems in conjunction with other software and communications technologies, are helping many companies turn information into knowledge by building interactive help systems.
There are six technologies that together are transforming the way some information center help desks operate. These same technologies will eventually transform customer support, sales support and emergency response centers, but it is at the Information Center Help Desk that they are finding the necessary level of tolerance for new technology introduction, albeit sometimes strained to the limit. These technologies are:
| Technology | Automates |
| Database | Record keeping |
| Expert Systems | Diagnostic Q & A |
| Decision Tree Software | Diagnostic Q & A |
| Text Retrieval | Search through manuals |
| Case-Based Retrieval | "Reminding" |
| Automatic Call Directing | Call routing |
All of these technologies, except for the last, are software technologies. In the near future, they will be joined by other technologies, for example voice-understanding and handwriting recognition, making for a very convenient user interface. But even now these systems work well and they are indicative of a new use of computers in large organizations. Until now, business systems have been primarily used to automate record-keeping tasks, from payroll to call tracking. In the future, the corporate computer network will become a new communications medium.
But let's return to the help desk. For a variety of reasons, it is not always clear which direction for automation a given Info Center Help Desk should choose. There are many vendors and their products are changing rapidly, often combining separate technologies into novel, integrated solutions. This article is a guide to the kinds of products available and to impact of the different technologies, based on our experience with help desks in Information Centers, customer support operations, and emergency response centers.
Some of the typical help desk problems, like call tracking and problem tracking, are best dealt with by conventional information systems. But many of the difficulties that help desk operations face are inherently knowledge problems. For example, a new TSR cannot use the information available to him or her in manuals, notebooks, databases and meetings without extensive training. Even experienced and talented TSR's have trouble integrating into their telephone troubleshooting the new problems that callers present daily and the fixes that the engineering department figures out.
Similarly, troubleshooting guidance is sometimes needed to keep up with new products, releases and repair procedures. This guidance is especially important when the TSR is trying to schedule properly equipped field service visits that will fix the customer's problem without a second trip. In addition to the troubleshooting task of figuring out what the customer is talking about, the TSR's behavior is often constrained to remain consistently in compliance with, for example, disaster recovery guidelines or a vendor's product return policy.
Moreover, even if a help desk operation is working well now, it won't forever. Key people leave, new systems get installed, and management decides to run it on a 24-hour basis. As with everything else in the business, automation is necessary because "if you don't, your competition will".
The inputs to the knowledge flow analysis are all the types of knowledge that TSR's must bring to bear in order to help a customer solve a problem. Such knowledge includes familiarity with the entire product line and the associated glitches, possible fixes, company policies and priorities in dealing with callers, and even knowledge about how the company's products interface with the products of other vendors. Also important is knowledge of how to use the existing documentation, database and textbase tools.
The analysis extends even further, to include all of the people in the knowledge chain. The callers have knowledge about their problem; the engineering department has knowledge about design decisions and changes, and product fixes. Marketing, training, documentation writers, field personnel and distributors also have relevant knowledge.
The output of the knowledge flow analysis is a determination of how the process could be improved: who could do a better job with better access to the relevant information. Once the critical knowledge problems are identified, the best knowledge distribution strategy can be designed. The solution may take the form of new documentation, communication strategy or training, or it may involve a knowledge system.
The help desk manager needs to keep an open mind about how information should be disseminated, and about the skill-levels for current jobs and the possibility of differentiating new jobs, which will be supported by the automation which is introduced. Besides the help desk operation, the documentation, marketing, communications, and training organizations should become familiar with the help desk technology and its implications to their functions. And hopefully so will the company's vendors, so that when the company gets a new product or new release, it can also get the product's troubleshooting knowledge base and plug it into the existing help desk system.
These products offer automation for the following kinds of activities:
Call Management: Answering and queuing incoming calls for the right TSR. Sometimes this is completely automated, so that the call is answered by a prerecorded voice and forwarded by the phone trees characteristic of automatic call directing (ACD) technology.
Customer tracking: What equipment does a customer have? Have they called before about this problem or other problems? Access control is also provided, and some systems even automate updating the billing database by TSR "connect time."
Help desk management.: Keeping track of what kinds of calls are coming in from whom, which ones get answered when, and how many callers each TSR handles per hour.
Product problem tracking: Which products are most troublesome and why? What are the unresolved customer problems, who is working on them, and which callers are waiting for answers?
On-line document retrieval: Pulling up the right page of the right document can be facilitated by new intelligent text retrieval, hypertext and CD-ROM technologies, including products like Verity's TOPIC, Knowledge Systems KMS, IBM's BookManager and many others. BusinessWise's SupportWise is a full-fledged call management tool with built in text-retrieval functionality.
The difficulty here is that for every problem there are a hundred ways to describe it, and sometimes text retrieval by keyword matching is not enough. These products use sophisticated text retrieval (e.g. parsing and synonyms), case-based reasoning (which structures the way cases are stored in the database), and even neural network technology (which mimics in software the pattern-matching ability of the human brain) to help find relevant cases. All these systems offer a call-tracking interface or fit in behind the one your TSR's are already using.
The simplest design is Bendata's HEAT, based on the results of the Help Desk Institute's consortium. On a hot key, behind a complete call-tracking workstation, is a topic outline that can be expanded by sub-headings or searched by keywords for relevant entries. The Mayflower Help Desk, Software Marketing Group's Help Express, Magic Solutions' Support Magic, Sheppard Systems' Help Desk, and Software Systems Corp.'s SOMADESK are a similar products with the ability to run on PC-LAN's. The latter three even offer prepacked "knowledge bases" about many of the most popular PC software, which you can buy and plug into your custom-built notebooks.
Answer Computer's apriori is a more sophisticated shared-notebook system which uses graphic user interfaces and sophisticated text indexing and retrieval schemes to allow TSR's to share information about current problems, how callers describe them, and what the fix is. The system actually learns, real-time, as TSR's are dealing with a call, and makes the information available, via networked UNIX workstations, to every TSR immediately. Old cases "decay" naturally over time.
Lysis Corp's Support Information System uses cased-based reasoning and sentence parsing technology on PC's to facilitate retrieval of relevant past cases. In this system, the TSR's do not enter information as free form text, but rather use "primitives" to develop a canonical underlying schema to uniquely capture problems and their descriptions.
CBR Express, from Inference Corporation, is built on top of an expert systems environment (see below), but is designed to be a case-based retrieval environment. CBR Express takes a text description of the problem and, using case-based algorithms, retrieves all the questions that might help pin down the problem. After answering them, another case-based search is performed, until the caller's problem is fixed.
Top of Mind, from the Malloy Group, is a novel PC-LAN system based on neural net pattern-matching technology to retrieve relevant case information. It has some parsing capability, like Lysis and CBR Express, and the "decaying reference" feature of apriori as well as a very nicely designed user interface. The metaphor of what comes first to mind is extended in this system to allow a fluid interaction about known problems.
Expert systems is a technology to help programmers build systems that contain a lot of knowledge. Unlike the products mentioned above, to use expert systems technology you have to commit a certain amount of effort to building and maintaining a program, called a knowledge base (KB). Sometimes this can be done by dedicated TSR's using decision trees or some other simple programming paradigm, and sometimes it requires full-fledged programmers. In either case, it is a major commitment of resources to getting the knowledge straight and encoding it into the software.
So why introduce an innocent help-desk operation to the tortures of software development and maintenance? In our experience, I have seen the following additional benefits of knowledge base systems, beyond those of call-tracking, text-retrieval and case-retrieval solutions:
Integrated with ACD. Perhaps the most novel customer support system product involving expert systems is IntelliSystems' TechSys. TechSys is designed to integrate with ACD technology to automate, completely, the "call screening and directing" task. In addition, the most common, "nuisance" problems can be diagnosed and handled by the system, over the phone, without TSR intervention. In the more difficult cases, the system helps the caller organize all the required information before passing the call on to a TSR. Among TechSys' installed base was the system which handled my distress call about Windows 3.0 this week. (The alternative to working with the automated system, I was told in its first pre-recorded message, was to wait approximately 30 minutes for the first available TSR.)
TechSys' knowledge base representation (see below) and total integration with ACD technology, voice mail, databases, billing systems, and the human technical support operation show it to be a well-thought-out system.
Special Purpose Shells and Overlays. Another feature of TechSys, which is common to all of the expert systems entries in the help desk arena, is that the knowledge representation scheme is specifically designed for customer support tasks. TechSys' KB, for instance, is organized around products, symptoms, and problems. A number of so called task-specific shells, designed for modeling support-related tasks like checklists and equipment diagnosis are now appearing. Some are designed for organizations already into expert systems, some for very complex diagnostic situations, while some use decision-tree interfaces to facilitate "end-user programming" of diagnostic procedures by the TSR's themselves.
One such tool is the Mahogany HelpDesk from Emerald Intelligence. Emerald's MahoganyPro tool was just another entry in the low-end shell game for PC's and Mac's, with some novel features for an inexpensive tool like point and click user interface, simple-to-use KB editing facilities, automatic integration of rule parameters with an object-oriented database, and pattern-matching rules over that database. But the Mahogany HelpDesk product goes well beyond that. It's HelpBuilder module is designed to support end-user development, that is, programming the KB by the TSR's themselves, by offering a decision-tree-like developer's interface.
Generally, these decision trees become quite awkward after the first 100 nodes, since new Q&A information must be placed in the right spot in the tree. Emerald is developing a technology for embedding trees within trees to handle the typical help desk situation where thousands of nodes may be required.
There are limits about how much programming to expect from the end users. The simple decision-tree development environment offered by the Mahogany HelpDesk and similar tools like Symbologic's Adept, TI's Procedural Consultant, and CAM Software's LogicTree, will only go so far. If the trees get too big, or if they need to be changed too often, or if they must be maintained by several different TSR's, they can become more of a nuisance than an aid. Furthermore, if data needs to be drawn in from a database, a programmer must get involved.
Beyond End-User Programming. Because the demands of the diagnostic task are only a subset of the capabilities of a full-fledged expert system development tool, many vendors are building front-ends to their tools to make it easier to build help-desk-type applications. They often then add some call tracking facilities to the user interface. One such overlay tool that sees itself being used by end users in conjunction with professional programmers, is the new Path Builder product from Aion, now in beta-testing. This task-specific overlay for Aion's best-selling ADS mainframe shell allows convenient representation of diagnostic decision trees. Path Builder can produce an ADS export file, which can then be integrated with other, programmer-produced ADS code using ADS's version maintenance feature.
Advantage KBS has gone one step further with their Help Desk Assistant. This task-specific help desk tool, again based on Aion's ADS, is designed for data center support operations. The system offers problem management facilities and integrates with IBM's Info/Man call-tracking product. What is novel about this offering is that you can buy some "knowledge modules" along with the tool. Currently modules for network problems and for PC problems are available. Obviously every data center will have their own knowledge about their own networks and PC's, but these knowledge products might be tremendous accelerators for organizations just starting to automate the help desk with knowledge systems.
Heavyweight Entries. It remains to be seen how far one can go with end-user programming for the help desk. In addition to the inherent complexity of large decision trees, there are attitudinal issues involving TSR's as programmers. And there are technical issues: Most importantly, how do you integrate a bunch of decision trees built by specialist TSR's on networked PC's into a logically consistent knowledge system offering all the relevant advice for every situation?
Whatever the eventual solution to this technical problem is, the end-user programming model will not cover all situations. In particular, when offering service aid for suites of complex equipment with tens of thousands of possible problems and rigorous diagnostic procedures, we are back in the realm of professional knowledge engineering. Several entries, for example, ROSH Intelligent Systems' CAIS, IntelliCorp's KLUE, AI Squared's IDEA, Automated Technology Systems I-CAT and Carnegie Group's Testbench, now offer workstation-based, task-specific tools for knowledge engineers faced with this level of diagnostic complexity.
Sold with Consulting. Several general-purpose shell vendors and systems-integration consulting firms now have service offerings specifically designed for help-desk applications, based on their experience, and often code, from other client's help-desk applications. The vendors specifically addressing help desks include Carnegie Group, IBM, AICorp, Bell Atlantic, Software Artistry, IntelliCorp, Andersen Consulting AI Squared and Pathfinder.
Netman, Computer Associates, San Jose, CA. Kristin Keyes, 408.432.1727.
SupportWise, BusinessWise, Campbell, CA. Mike Cohn, 408.866.5960.
Topic, Verity, Mountain View, California. Robert Williams. 415.960.7600.
CBR Express, Inference Corporation, Los Angeles, CA. Luis Sastre, 415.552.3610.
HEAT, Bendata Management Systems, Richardson, Colorado Springs, CO. Lee Lowry, 719.531.5007.
The Help Desk, Sheppard Systems, Houston, TX. Dorian Cameron, 800.888.6534.
Help Express, Software Marketing Group, Des Moines, Iowa. Doug Nicholas, 800.395.0209.
Mayflower Help Desk, Mayflower, Burlington, MA. Steve Harvey, 617.270.9000.
SOMADESK, Software Systems Corp., East Brunswick, NJ. Len Bonfiglio, 908.613.8100.
Support Information System, Lysis, Inc., Decatur, Georgia. 404.373.3359.
Support Magic, Magic Solutions, Mahwah, NJ. Vincent Van Dyke, 201.529.5523.
Top of Mind, The Malloy Group, Morris Hills, NJ. Peter Dorfman, (201)267.4464.
LogicTree, CAM Software, Provo, Utah. 801.373.4080.
Mahogany HelpDesk, Emerald Intelligence, Ann Arbor, Michigan. David Bowers, 313.663.8757.
Procedural Consultant, TI Express, Houston, TX. Herb Roehrig, 713.894.3940.
Help Desk Assistant, Advantage KBS, Edison, New Jersey. 201.287.2236.
CAIS Adviser, ROSH Intelligent Systems, Needham, Massachusetts. Elan Blair, 617.449.0049.
KLUE, IntelliCorp, Mountain View, California. 415.965.5500.
Testbench, Carnegie Group, Inc., Pittsburgh, PA. Chip Bodin, 412.642.6900.
I-CAT, Automated Technology Systems, Hauppauge, NY. David Butland, 516.231.7777.
AI Squared, North Chelmsford, MA. Kevin Flood, 508.250.4000.
AIM Consulting, Pleasant Hill, California. Keith Telle, 415.944.4749.
Andersen Consulting, Chicago, Illinois. Joe Carter. 312.580.0033.
Bell Atlantic, Princeton, New Jersey. 609.275.4545.
Carnegie Group, Pittsburgh, PA. Florence Hendersen. 412.642.6906
Crescent Project Management, Palo Alto, California. Jacky Hood, 415.328.1720.
The Help Desk Institute, Colorado Springs, Colorado. 800.248.5667.
IBM, Palo Alto, California. 415.855.4076.
Pathfinder, Jacksonville, Florida. Robert Touchton. 904.296.1685.
Software Artistry, Indianapolis, Indiana. 317.876.3042.
